The researchers “use de-identified smartphone geolocation data for a sample of US phones from January 2019 to February 2020,”…
The researchers “use de-identified smartphone geolocation data for a sample of US phones from January 2019 to February 2020,” obtained “from an online data vendor that provides data commercially to businesses, governments, and researchers” and “works with numerous mobile application providers that track ‘pings’ of the location of a phone while the application is either currently in use or is running in the background.” Then they use the addresses of SEC offices and corporate headquarters, and then match the smartphone pings to the buildings. A smartphone is assumed to belong to an SEC employee if it “pinged for at least 20 unique workday hours within one SEC location during the month” and “the accumulated time in that SEC building [is] greater than in any other buildings in the respective month.” And then they go measure which companies those SEC employees visited.
Matt Levine casually mentioning this in an aside like it’s normal